TranslatorQformerArxiv类的forward函数中注释image-text contrastive是什么意思？

作者你好，
请问`/Translator/models/translator_models/translator_qformer_arxiv.py`中，`forward`函数的注释image-text contrastive是什么意思？arXiv的数据集节点好像也没有image，是指graph node与text对比吗？
还有第一阶段的Train中计算$s_{positive}$和$s_{negative}$的代码在什么位置。
望回复，万分感谢

Start with /Translator/models/translator_models/translator_qformer_arxiv.py and inspect the forward function's “image-text contrastive” comment in context. Then search the repository for s_positive and s_negative and the first-stage training code. Done means documenting what the comment refers to and identifying where those two scores are computed.

Written by the indexing model from the issue text.